Frames and Features Selection

Table 5.1 lists the two competing basic discourses and their constituent frame signifiers that were invoked strategically by political elite actors in the context of intelligence-gathering to combat terrorism. The securitizing basic discourse emphasized the need to collect information using any means necessary. It was expressed by the executive branch through an interrogation frame that justified, protected and promoted the

successes of its CIA program. Whether described as “harsh interrogation” or “enhanced interrogation techniques”, the interrogation frame was employed to legitimize

exceptional measures in the name of security. The Bush administration also

downplayed the exceptionalism of torture in an attempt to legitimize it (Rowling et al., 2011). A desecuritizing discourse, conversely, sought to delegitimize the CIA program through the torture frame. Adherents of this desecuritizing discourse used the torture frame to argue that the program constituted torture and ran counter to both American and international norms, laws and values. It was employed to scale back the

exceptionalism of the threat: while terrorism remained a threat, it was not sufficient to overstep these institutions. These two issue frames formed different schematic

interpretations; like Chong and Druckman’s (2007) study on reactions to a white supremacist rally, the two frames on intelligence-gathering here became associated with different signifiers – torture with human rights and liberal values considerations, and interrogation with security and protection against the threat of terrorism.

Table 5.1 Content Analysis Frame Signifiers

Remedy Proposal/Issue Frame

Securitizing discourse Interrogation

Desecuritizing discourse Torture

These signifiers, torture and interrogation, are thus used to drive the content analysis.

While the discourse analysis demonstrated that each signifier was associated with a particular broader frame – and thus activated different schematic interpretations by the audience – previous research has also demonstrated differential effects. Rios and Mischkowski (2018), for example, demonstrate through a series of five framing experiments that using the signifier torture produced more negative evaluations than interrogation. Similarly, Blauwkamp, Rowling and Pettit’s (2018) experimental study demonstrates that framing in poll questions influences torture preferences – but they do not link attitudes to political and media frames. The approach here empirically tests whether similar framing effects emerge in natural settings where political elites and the press supply the information marketplace with competing discourses.

Existing research on framing by press and political elite actors has also explored competing frames by using signifiers. Bennett et al. (2006) distinguish between abuse and torture to demonstrate that the indexing model best reflected the relationship between news media and political elites. In their eight-month content analysis from 2004, they find that the press largely adopted the Bush administration’s preferred abuse frame over the less popular torture frame. Rowling et al. (2011) expand on this study and examine network and print news actors as well as congressional and executive texts over a longer period (2004-2006), to conclude – contra Bennett et al. (2006) – in favor of the cascading activation model. In short, they find that while congressional

Democrats employed the torture frame and challenged the Bush administration’s preferred frames, the media neglected these cues. Jones and Sheets (2009) similarly find that the abuse frame dominated the torture frame in American print media outlets from 2004 to 2005.

While these studies are individually informative to understanding the competitive framing of torture, their confinement to the few years surrounding Abu Ghraib predates shifts in public opinion toward support for torture, and overlooks long-term trends in discourse, like the shift toward the interrogation frame and contextual political changes like the transition to the Obama administration. Further, the previous studies have restricted content analysis to certain network and print media actors – cable news has been largely ignored. This is a glaring omission given that cable news commands the widest audience in the American media landscape. Finally, extant literature has yet to link frames in natural settings (as opposed to experimental settings) – whether from the press or political elites – to attitudes on torture. The remainder of this chapter addresses these gaps.

One newspaper – the New York Times – and three television networks – CNN, Fox News and MSNBC – are analyzed, collectively representing a spectrum of ideology and mediums. The broad corpus of texts was downloaded from the Nexis Academic

database for each of the four news sources using the following search query:

terrori! OR gwot OR "war on terror" OR "overseas contingency operation"⁴⁵

Political elite content was retrieved using similar search queries from the American Presidency Project and the Congressional Record. The keywords in the search query accommodate the different ways to represent the terrorism threat frame, aimed at

casting a wide net and capturing all relevant articles. For all sources, the data was subset to content published between January 1, 2001, and December 31, 2016 – as discussed in Chapter 3, this timeframe was chosen to encompass the entire Bush and nearly entire Obama administrations. The discourse analysis in the previous chapter revealed that both administrations took different positions in the torture debate, ensuring that this selected timeframe will capture rich variability in official framing. Guided by Entman’s cascade activation model and framing theory’s expectation that downstream actors are

45 Items in quotes are searched as a phrase; exclamation marks indicate wildcard searches, where terrori!

matches “terrorism”, “terrorist” and “terrorists”. The term “gwot” is shorthand for “global war on terror[ism]”.

bounded in their range of discourse by cues from political elites, this variability is assumed to extend to the press.

Table 5.2 lists the total number of texts collected and analyzed for each source, which together total 254,250 – this includes news articles for the New York Times; transcripts of individual shows for each of the television networks; texts from the Congressional Record; and speeches, press briefings, remarks, debate transcripts, addresses, and a variety of other texts attributed to the president and vice president. Unsurprisingly, news coverage largely exceeds political discourse in number of texts – as noted previously, news coverage is more voluminous while political discourse is more symbolic. Notably, however, the number of MSNBC segments that mentioned terrorism is only slightly more than the number of executive branch texts for the same period and less than the number of congressional texts. This may be attributed to MSNBC’s domestic

orientation: a study led by the Pew Research Center found that CNN’s coverage of foreign stories (30% in 2007; 23% in 2012) led both Fox News (21% in 2007; 15% in 2012) and MSNBC (25% in 2007; 7% in 2012)⁴⁶ – findings that are fully consistent with the data in table 5.2 (Jurkowitz et al., 2013).

Table 5.2 Number of texts by source

Source Number of texts

CNN 79,145

Fox News 31,172

MSNBC 10,580

New York Times 101,782

Presidential Elites 7,031 Congressional Elites 24,558

Total 254,250

46 While the Pew study (Jurkowitz et al., 2013) does not specify which topics were included in “political”

or “foreign” stories, the attribution of terrorism to foreign agents – e.g., al Qaeda and Osama bin Laden – and the protection against threats from abroad suggests that coverage of terrorism would be contained in

“foreign” news.

5.2.3 Frame Competition 5.2.3.1 Aggregate Analysis

Given the wide consensus in literature that framing effects are primarily contingent on the frequency of exposure to a particular frame (e.g., Domke et al., 1999; Entman, 2004; Oren & Solomon, 2015; Price & Tewksbury, 1996), frame competition is

operationalized here by comparing the prevalence of the securitizing and desecuritizing frames within each source. A Python script was used to iterate over all texts, counting the occurrences of specific keywords and patterns that were linked to each of the two frames (see Appendix A for details on frame features and extraction methods). Table 5.3 shows the frequency of occurrences of the securitizing (interrogation) and

desecuritizing (torture) frames, and for additional context, the threat (terrorism) frame.

Despite the Times having a higher aggregate volume of texts in the complete corpus (table 5.2), CNN had the highest frame count across all three frames – this is primarily because television news segments are longer than print news articles⁴⁷ and thus contain potentially more frame instances. The relative primacy of CNN across all three frames in terms of total counts thus reflects not only its relatively large word count, but also its stronger foreign-policy calibration among the different cable news actors. The length and political commentary format of television shows also allows for a more diverse set of topics which explains the low frame rates (measured as frame occurrence per-ten-thousand words). Newspaper articles, conversely, are focused on specific topics within each article, yielding a higher frame rate. For the Times then, when a frame was

present, it was more likely to be the focus of the entire article.

47 The average word count for each source was: 5,302 for CNN; 2,911 for Fox News; 8,256 for MSNBC;

852 for the Times; 2,226 for presidential elites; and 8,397 for congressional elites.

Table 5.3 Terrorism and Issue Frame Aggregate Prevalence by Source

Terrorism Interrogation Torture CNN

Total 315,838 15,803 17,466

Rate 7.526 0.377 0.416

Fox News

Total 113,350 6,632 5,473

Rate 12.491 0.731 0.603

MSNBC

Total 61,112 4,485 9,069

Rate 6.997 0.513 1.038

New York Times

Total 158,527 10,612 11,571

Rate 18.291 1.224 1.335

Presidential Elites

Total 34,863 452 714

Rate 22.334 0.290 0.457

Congressional Elites

Total 198,741 6,536 9,436

Rate 9.638 0.317 0.458

All Sources

Total 901,209 44,945 54,577

Rate 9.774 0.487 0.592

Note: Total values indicate the number of instances of each frame. Rate is calculated as the presence of the frame per ten-thousand words.

Turning to the two opposing issue frames, the ratios of securitizing frames to

desecuritizing frames in figure 5.1 suggest that, with the exception of Fox News, actors were likelier to employ the desecuritizing torture frame. Fox News alone was more likely to employ the securitizing interrogation frame – this is unsurprising given that it consistently provided airtime to Bush administration officials. Its conservative and Republican orientation drove it toward the preferred party frame and actors.

Interestingly, figure 5.2 shows that Fox News leaned slightly toward the desecuritizing frame throughout the Bush administration; its aggressive securitizing strategy occurred largely in the Obama administration – in the absence of Republican political cues –

where it led with the securitizing frame by a factor of 1.7 to one. While this rebellion against Obama’s preferred anti-torture stance presents strong evidence of at least one press agent acting independently of the executive branch, it does tentatively suggest an alignment between Fox News and Congress, whose coverage also became increasingly securitizing during the Obama administration according to figure 5.2.

Figure 5.1 Aggregate Ratio of Securitizing Frame to Desecuritizing Frame (2001-2016)

Note: Ratio is calculated by dividing the number of securitizing frames by the number of desecuritizing frames. The dashed line represents the cutoff for equal coverage of both securitizing and desecuritizing frames.

Figure 5.2 Ratio of Securitizing Frame to Desecuritizing Frame by Presidential Administration (2001-2016)

Note: Ratio is calculated by dividing the number of securitizing frames by the number of desecuritizing frames.

At the other end of the spectrum, MSNBC covered the securitizing frame least among press actors in volume, and least among all actors relative to the desecuritizing frame.

At the aggregate level, it covered the torture frame over twice as much as the

interrogation frame. This performance was consistent across both administrations, and unlike Fox News, MSNBC’s coverage shifted in the opposite direction toward more desecuritizing coverage from the Bush years to the Obama years. The Times also moved toward increased desecuritizing coverage, but at a steeper rate than MSNBC, indicating a more drastic shift in tone between the two administrations. While the Times’ ratio of coverage is similar to CNN’s at the aggregate level, figure 5.2 show a critical difference: CNN’s coverage did not change very much, and remained fairly balanced through both presidents. In this regard, it shared with MSNBC a resistance to shift its tone between different political contexts, also indicating some separation from official influence on coverage between both administrations.

The wide disparity between Fox News and MSNBC suggests that these two sources are likely to have strong framing effects on their audiences in opposite directions. Framing theory suggests that Fox News audiences, more exposed to the securitizing frame, are more likely to support torture as an exceptional measure, whereas MSNBC viewers, predominantly exposed to the desecuritizing frame, are likelier to reject torture. This expectation is strengthened by the increased coverage both actors gave to their preferred frame over time: figure 5.3 shows specifically that Fox News covered the securitizing frame 19% higher in the Obama years than in the Bush years, while MSNBC covered the desecuritizing frame 14% higher in the same timeframe. This is striking given that both decreased their coverage of the terrorism frame by 44% (Fox News) and 60% (MSNBC). Despite declining coverage among all other actors of all three frames, Fox News and MSNBC alone increased coverage of their preferred frames. Like MSNBC, the Times’ shift towards the desecuritizing frame suggests a likely alignment between its readers and opposition to torture during the Obama administration. On the other hand, CNN appears less likely to produce framing effects given its ambivalent coverage of both frames (see Druckman, 2004).

Figure 5.3 Percent Change in Frame Volume Between Bush and Obama Administrations

Note: For each frame and actor, percent change is calculated by subtracting the number of invocations during the Bush administration from the number of invocations during the Obama administration and dividing by the number of invocations during the Bush administration.

While at the aggregate level, both congressional actors and White House leaders were more likely to employ the desecuritizing frame, the differences between administrations show surprising results. Both congressional elites and presidential elites increased their coverage of the securitizing frame relative to the desecuritizing frame. The latter finding is especially interesting, because it suggests that the Obama administration was more likely to use the securitizing frame than the Bush administration, even though both showed an overall preference for the desecuritizing frame. This largely reflects the impact of Abu Ghraib, which necessitated a strong and loud anti-torture stance from the Bush White House (as discussed in chapter 4). During this phase, the Bush

administration invoked the desecuritizing frame much more aggressively than either administration did in later phases. Moreover, Obama’s general disengagement from the debate – and thus fewer invocations of either frame evidenced in figure 5.3 – makes the ratio more sensitive to small changes. Congress, similarly, shifted from a heavy focus on the torture frame in the early post-Abu Ghraib period toward mostly securitizing coverage in the Obama years. But like presidential elites, Congress also invoked both frames much less, making its ratio of securitizing to desecuritizing more sensitive to small differences between the two. The drastic drop in coverage for both sets of

political elites is illustrated in figure 5.3 – both Congress and presidential elites showed the highest drop in coverage of all three frames across all actors, suggesting that the press largely kept terrorism part of public discourse.

5.2.3.2 Quarterly Time-Series Analyses

Disaggregating the data into quarterly time-series illustrates further nuances in frame invocations over time. Figure 5.4 shows the volume of securitizing and desecuritizing frames, measured in occurrences per ten-thousand words. It shows how much of overall terrorism discourse was made up of the torture and interrogation frames for each actor.

Higher scores indicate periods of increased salience of either frame in terrorism

discourse. Figure 5.5 illustrates the competition between the two frames: specifically, it plots the proportion of securitizing frames that occur in the sum of both securitizing and desecuritizing frames. In each panel of figure 5.5, the solid horizontal line at y=0.5 marks the point at which the volume of securitizing and desecuritizing would be equal:

the space above the line indicates a louder securitizing frame, while the space below indicates a louder desecuritizing frame. The vertical dashed line in the middle of each panel in both figures 5.4 and 5.5 marks the point at which the Obama administration came into office.

Figure 5.4 Frame Coverage as Word per Ten-Thousand for News Media and Political Elites

Note: Each panel plots the rate of the interrogation and torture frames per ten-thousand words for each source. The vertical dashed line in 2009 marks the beginning of the Obama administration.

Figure 5.5 Quarterly Interrogation-Torture Proportion (2001-2016)

Note: Each panel plots the proportion of securitizing frames that occur in the sum of both securitizing and desecuritizing frames for each source. The area above the horizontal line at y = 0.5 indicates a louder securitizing frame. The vertical dashed line in 2009 marks the beginning of the Obama administration.

Gaps in the presidential discourse plot indicate periods of no coverage of either frame.

The results are congruent with aggregate findings, but also provide critical insight into frame competition and salience in different political contexts. Beginning with press actors, figure 5.4 suggests that CNN was largely balanced at the aggregate level, but figure 5.5 suggests that its coverage was cyclical – while it led with the interrogation frame during the Bush administration’s first term, it mostly invoked the desecuritizing in the second and similarly oscillated throughout the Obama administration. Compared to other press actors, its coverage of both frames was less salient in overall terrorism discourse (measured in frame coverage per ten-thousand words). Both its ambiguous coverage and low salience of either frame suggests it likely had little if any effect on public attitudes toward torture. The Times, on the other hand, largely invoked the desecuritizing frame beginning in 2010 after a tenure of mixed coverage during the Bush administration. It shared with MSNBC a preference for the desecuritizing frame

during Obama’s administration, though MSNBC overwhelmingly reflected this pattern across almost all quarters, as evidenced by figure 5.5. During Obama’s presidency, only Fox News maintained a strong securitizing frame (relative to the desecuritizing frame). Figure 5.4 shows periods of high interrogation invocations by Fox News in multiple quarters during the Obama presidency – notably, many of these peaks are not echoed by other media actors. Again, Fox News was the only actor that invoked the securitizing interrogation frame more during Obama’s presidency than during Bush’s, while MSNBC was the only actor that increased its number of invocations of the desecuritizing torture frame from Bush to Obama.

The time-series data also show that presidential elite content is mostly consistent with the discussion in the discourse analysis: the desecuritizing frame was adopted in the aftermath of Abu Ghraib (2004) during which the Bush administration insisted that torture was not parts of its official policy; rebranding and justification followed with a brief preference for the securitizing frame; and finally, an overall silencing of the debate occurred throughout Obama’s presidency, as indicated by several gaps in figure 5.5. Still figure 5.5 provides evidence that Obama engaged the interrogation frame at a higher rate than expected – spikes in 2009 and 2014 correspond to Obama’s executive order (which banned torture) and the release of the Senate Intelligence Committee Report on the CIA’s program. In several quarters, Congress also gave more attention to the securitizing frame than the desecuritizing frame. Again, these peaks of securitizing coverage by congressional and presidential elites reflect an increased sensitivity given the reduced usage of both the securitizing and desecuritizing frames between the Bush and Obama presidencies.

These results reaffirm previous findings of the media acting independently of political elites. MSNBC and the Times both moved toward increased relative attention to the desecuritizing frame from one presidential administration to the next, while political elites – both Congress and the executive branch – moved in the other direction. The

These results reaffirm previous findings of the media acting independently of political elites. MSNBC and the Times both moved toward increased relative attention to the desecuritizing frame from one presidential administration to the next, while political elites – both Congress and the executive branch – moved in the other direction. The